Brainer

The Brainer provides a uniform graphical user interface and standard procedures for numerous machine learning methods, like Artificial Neural Networks, Support Vector Machines, Classification and Regression Trees, and many more as well as classical regression models.

Beside the standard procedures for training, validation of various machine learning models a lot of different methods for the evaluation, assessment and rating of the models are available.

Additionally meta-modelling methods are provided, like building of commitees, automatic selection of the best input parameters, and many more.

Selected models can directly be exported from Brainer to various targets platforms.

In principle Brainer is designed open, enabling to implement custom made model types, training procedures, evaluation, and rating functions. That way the Brainer can also be used for the automated calibration of classical models.

• Import of different data formats
• Evaluation of data from distributed sources
• Application of numerous data analysis methods
• Data preparation (e.g. PCA, ICA,...)
• Different, customizable methods for manual and automated data transformation for data normalization
• Execution of training and validation of numerous machine learning models
  • Artificial Neural Networks
  • Support Vector Machines
  • Classification and Regression Trees
  • ...
• Conflict analysis for the identification of inconsistent model parameters and training data
• Building of commitees
• Several methods for automated feature selection
• Direct export of the models to C, FORTRAN, Python, m-Code, XML parameter files,...
• freely configurable plots and visualization methods for the model evaluation and rating
• Robustness evaluation of the models
• ...

Technically the Brainer is a MATLAB-Toolbox, which is built on top of the

• Deep Learning Toolbox and the
• Statistics and Machine Learning Toolbox

from The Mathworks Inc. In addition also other, external models, training procedures, rating functions and evaluation procedures can be integrated for customization.

Beside many common formats especially time series and signal data from the Stipulator can be taken directly.

All functions of the Brainer can also be controlled by a command line interface. That way application specific procedures can be composed into custom scripts and apps for process standardization and customer specific integration in arbitrary custom environments.

MATLAB is a registered trademark of The Mathworks, Inc.